Diagnostic Verification of Temperature Forecasts

Publisher Website
Google Scholar
Abstract
A diagnostic approach to forecast verification is described and illustrated. This approach is based on a general framework for forecast verification. It is “diagnostic” in the sense that it focuses on the fundamental characteristics of the forecasts, the corresponding observations, and their relationship. Three classes of diagnostic verification methods are identified: 1) the joint distribution of forecasts and observations and conditional and marginal distributions associated with factorizations of this joint distribution; 2) summary measures of these joint, conditional, and marginal distributions; and 3) performance measures and their decompositions. Linear regression models that can be used to describe the relationship between forecasts and observations are also presented. Graphical displays are advanced as a means of enhancing the utility of this body of diagnostic verification methodology. A sample of National Weather Service maximum temperature forecasts (and observations) for Minneapolis, ... Abstract A diagnostic approach to forecast verification is described and illustrated. This approach is based on a general framework for forecast verification. It is “diagnostic” in the sense that it focuses on the fundamental characteristics of the forecasts, the corresponding observations, and their relationship. Three classes of diagnostic verification methods are identified: 1) the joint distribution of forecasts and observations and conditional and marginal distributions associated with factorizations of this joint distribution; 2) summary measures of these joint, conditional, and marginal distributions; and 3) performance measures and their decompositions. Linear regression models that can be used to describe the relationship between forecasts and observations are also presented. Graphical displays are advanced as a means of enhancing the utility of this body of diagnostic verification methodology. A sample of National Weather Service maximum temperature forecasts (and observations) for Minneapolis, ...

This publication has 0 references indexed in Scilit: